Analysis of Microarray Data Using Z Score Transformation Chris Cheadle, Marquis P. Vawter, William J. Freed, Kevin G. Becker  The Journal of Molecular Diagnostics  Volume 5, Issue 2, Pages 73-81 (May 2003) DOI: 10.1016/S1525-1578(10)60455-2 Copyright © 2003 American Society for Investigative Pathology and Association for Molecular Pathology Terms and Conditions

Figure 1 A portion of the hybridization images of three NIA-Immunoarray filters hybridized to radiolabeled total RNA from three different biological conditions (untreated T cells, control; PMA plus ionomycin, PMA+I; and PMA plus anti-CD28, PMA + 28). The grid used for quantitation is shown superimposed on each image. The increase in gene expression of interferon-γ (IFNgamma) is shown. The Journal of Molecular Diagnostics 2003 5, 73-81DOI: (10.1016/S1525-1578(10)60455-2) Copyright © 2003 American Society for Investigative Pathology and Association for Molecular Pathology Terms and Conditions

Figure 2 Scatter plot showing the linear relationship between measurements of changes in gene expression using either globally normalized data (fold change) or Z transformed data (Z ratio) on a lognormal scale. The r2 value of the linear regression = 0.961. The Journal of Molecular Diagnostics 2003 5, 73-81DOI: (10.1016/S1525-1578(10)60455-2) Copyright © 2003 American Society for Investigative Pathology and Association for Molecular Pathology Terms and Conditions

Figure 3 A–C: Scatter plots of treated PMA plus ionomycin, PMA+I; D–F: PMA plus anti-CD28, PMA + 28; untreated cells, control (all panels). A and D compare the raw data. B and E compare globally normalized data. C and F compare Z score data. The distortion in the data (E) resulted in 226 genes being called as up-regulated by global normalization in the PMA + 28 treatment. As shown for Z score transformation, this distortion is largely reduced (F) resulting in 24 genes being up regulated. Arrows indicate areas of data distortion introduced by the PMA + 28 hybridization filter. The Journal of Molecular Diagnostics 2003 5, 73-81DOI: (10.1016/S1525-1578(10)60455-2) Copyright © 2003 American Society for Investigative Pathology and Association for Molecular Pathology Terms and Conditions

Figure 4 Cluster analysis of Z score data, sorting both genes and experiments simultaneously. Shown here is data from control cells only. Samples include microarray data from both Jurkat as well as donor PBTs (D1, D2, D3). RNA labeling replicates are alphabetical (A, B, C), duplicate data from the same filter are numeric(1,2). Scale of red to green equals higher to lower gene expression. The Journal of Molecular Diagnostics 2003 5, 73-81DOI: (10.1016/S1525-1578(10)60455-2) Copyright © 2003 American Society for Investigative Pathology and Association for Molecular Pathology Terms and Conditions

Figure 5 Comparison of several different methods for estimating significant changes in gene expression (highlighted by shaded values) using the same dataset. Data shown is a sampling of the most significant genes up-regulated by treatment with either PMA plus ionomycin (A) or PMA plus an antibody to the CD28 receptor (B). Significance thresholds for fold change were set at ≥ 2, for Z ratios at ≥ 1.5, for Z test at P ≤ 0.01, for SAM at d > 1.45. SAM, fold changes, Z ratios, and Z test data are from a single donor. Labeling replicates (3 for the control sample RNA, 2 each for PMA+I and PMA + 28 sample RNAs) were first averaged for Z ratio and fold change estimates. These results are compared to a composite of replicated data (using averages of multiple individual donor RNA labelings) from three individuals (all donors Z test). The data are sorted in order of decreasing Z test value for the all donors column. The Journal of Molecular Diagnostics 2003 5, 73-81DOI: (10.1016/S1525-1578(10)60455-2) Copyright © 2003 American Society for Investigative Pathology and Association for Molecular Pathology Terms and Conditions